Back to the drawing board: assessing siting guidelines for sand dams in Kenya

Arid and semiarid lands (ASALs) makeup 40% of the global land area and are occupied by 2.3 billion people, including 74% of the world’s poorest (Mortimore et al. 2009). Limited access to water is to blame for many problems and disasters associated with ASALs. Poor rainfall characteristics result in poor livelihoods. In Kitui, Kenya, for example, recurrent crop failure affects access to food and incomes at least once every five seasons (Omoyo et al. 2015). In most other seasons, crops suffer suppressed growth due to less than adequate rains, but also because unreliable and unpredictable rainfall hinders investment in inputs for farming. This has led to a high poverty index of 63.5% in Kitui County against the national average of 47.2% (KCRA 2012).

As many others in ASALS, Kitui’s inhabitants spend much of their time searching for water (Thomas 2000). Water sources in Kitui County include boreholes, shallow wells, earth dams, roof rainwater harvesting, seasonal rivers and sand dams. Boreholes are expensive to develop requiring support from the county government or donors. Furthermore, Kitui’s boreholes often run dry after (Ngigi 2003) sometime due to limited aquifer recharge (Mburu 1989). A high number of boreholes in Kitui County have brackish water due to mineral composition of the parent rocks, especially those to the south of the county (Munyao et al. 2004). Shallow wells are not reliable either. They run dry as the drought intensity increases due to the receding water table. Small size earth dams are unsuitable as water sources due to the very high potential evaporation rates of between 1800–2400 mm per year (CGK 2014). Larger earth dams are expensive and generally all earth dams require regular desilting which is expensive. Roof rainwater harvesting systems are expensive and are practiced by only a few people who can afford them.

As such, many of Kitui’s potential water sources are known to have problems, but so far the general opinion appears to be that sand dams (also known as sand-storage dams), built to increase water access from sandy seasonal rivers, are generally highly successful and a major technology to be implemented elsewhere (Thomas 2000; Lasage 2008; Quilis et al. 2009; Stern and Stern 2011). Storing water in sand dams has advantages compared to conventional dams. For example, sand dams reduce water evaporation substantially and do not result in breaching that could cause catastrophic accidents (Mansell and Hussey 2005). Beyond 60 cm below the sand level, evaporation becomes negligible (Wheater and Al-Weshah 2002; Diettrich, 2002; Sivils and Brock, 1981; Stern and Stern, 2011; Ertsen and Hut 2009). They minimize water contamination through livestock and wildlife droppings and human use (Nissen-Petersen, 2000). They do not result in breeding of mosquitoes (Nissen-Petersen 2006) and do not take up valuable land (Ertsen et al. 2005).

Typically, to guarantee success, sand dam sites are selected with certain criteria, that are considered to contribute to sand dams’ ability to increase water supply from a riverbed. This paper explores whether existing criteria to decide where to locate the dams are adequate guides to ensure the sand dams provide water for community use during drought periods as evidenced by the presence of water during the period of the assessment as well as community interviews on when they usually stopped obtaining water from the sand dams. The paper also assesses whether reliance on such guidelines should be revised or not. Factors that were assessed include; soil, rainfall amount, topography, geology, presence of sub-surface water on a section of the riverbed for about a month after it got depleted in other sections and water indicating vegetation. This paper explores these environmental factors and their complexity as often they are dynamic and unpredictable. The factors were assessed individually to identify patterns that could be attributed to each. Assessing the concept of agro-ecological zones (AEZs) allowed considering possible interactions of these factors (Sombroek et al. 1982) and their combined influence in determining improved access to water from sand dams. An additional factor of stream order as proposed by (Strahler 1957) was also included in the assessment.

This study looks into the past to ascertain whether the forecasting of these criteria in Tiva catchment in Kitui County was accurate. Sand dams have indeed been successful in Kitui, in terms of increased access to water (Mansell and Hussey 2005), higher incomes (Lasage 2008; Excellent Development 2015), vegetation rejuvenation (Manzi and Kuria 2011), recharging groundwater (Thomas 2000), and in some cases rejuvenated streams (Ngigi 2003). Our paper suggests, however, that any unilateral success of sand dams may not be found. Even when constructed on locations that meet all the recommended environmental factors, many sand dams constructed along Tiva catchment in Kitui’s County failed to provide water for community use as expected. We suggest that using sand dams to increase water access should be approached with more caution than has been the case in the past. Our findings are important for the larger debate on water shortages faced by communities in the ASALs and consequent implementation of new technologies for improving water access.

The number of sand dams that exist in the world may not be known. Excellent Development quotes about 4000 globally (Excellent Development 2015) with about 50% constructed in South Eastern Kenya counties, but the method used to establish this number is not indicated. As such, it is very likely that South Eastern Kenya has the largest concentration of sand dams on the globe, all developed in the last 100 years (Excellent Development 2015), with the majority from 1990s onwards (De Trincheria et al. 2015). These impressive numbers notwithstanding, we argue that similar to many other water resources development efforts in arid and semi-arid regions, sand dams face sustainability challenges due to uninformed decisions that are not backed by locally acquired data (see Wheater and Al-Weshah 2002). Sand dams bring renewed hope to communities who may even consider expanding their resource base to improve their livelihoods. Consequently, the failure of these interventions to improve access to water in ASALs is a problem that needs to be addressed. Lack of water adversely affects access to other basic necessities resulting in (a danger of) chronic poverty (Republic of Kenya 2015; Ngigi 2003).

In the sections that follow, sand dams are introduced. Then the method used in this study is outlined in more detail, followed by results and their discussion. Since sand dams are relatively expensive to construct given the high poverty levels in Kitui (MOA 2014), the conclusions and recommendations guide on what precautionary measures to use to minimize wastage of resources when building sand dams. Generally, this paper suggests that sand dams’ designers and stakeholders should go back to the drawing board as reliance on sand dams to provide water during the drought periods needs to be re-evaluated.

As elsewhere, Kenya’s ASALs (80% of its area) are fragile environments, characterized by little and erratic, often unreliable rainfall (Shisanya, et al,. 2011). Tiva River catchment in Kitui County is no exception to this. The area receives two rainfall seasons per year: the short and long rainy seasons, which occur in November–December and April–May, respectively, with two inter-seasonal drought periods. The rainy seasons suffer wide variations in rainfall amounts and distribution from year to year (Lasage 2008). Rainfall failure in both seasons is common. The higher areas around Kitui town receive between 500 and 1050 mm (GOK 2013). Amounts decline to the South, with annual totals of less than 500 mm. With average temperatures relatively high at 24 °C, potential annual evapotranspiration in Kitui County is 1800–2000 mm (Kuria et al. 2012; Munyao et al. 2004). The inter-seasonal drought of June to October is more severe than the one from January to March during which the land receives occasional rain showers.

Apart from Tana and Athi, all other rivers in Kitui County are seasonal. Their riverbeds contain water for a limited period during and immediately following the rainy season. However, the seemingly dry riverbeds may carry subsurface water beneath the sand for prolonged periods of time after the rainy season. Subsurface flow rates can be substantial. A study in Zimbabwe that assessed flow rates in four sandy rivers estimated about 100m³ per day (Mansell and Husseys 2005). It is this subsurface water that has been used by Kitui’s communities since time immemorial. The local people scoop out the sand from the seemingly dry riverbeds to reach for water which lies beneath. As the drought period progresses, the water level in the sand bed drops, sometimes to a few meters deep (Mburu 1989), making access to water from the sandy riverbeds difficult and dangerous. Increasing access to water from these riverbeds is important as they are the most widely distributed water source in Kitui County.

To increase water volumes from these riverbeds, sand dams and to a lesser extent subsurface dams were mobilized. Sand dams are impermeable barriers (such as the reinforced masonry barrier in Fig. 1a and b) erected across the seasonal riverbeds to impound more sand and thus water (which comprises up to about 40% of the sand volume) (Mburu 1989; Thomas 2000). Unlike sand dams, sub-surface dams are similarly built but do not rise above the original sand level (Ngigi 2003). Sometimes sand dams are fitted with a shallow well on the upstream side from which water will be drawn. As a specific type of groundwater dam, sand dams have been a common feature in many arid parts of the world such as the Middle East, China, South Africa, Zimbabwe, Mozambique, India, Japan and United States of America (USA) (Diettrich 2002). In the USA, sand dams were used to provide water in the South Western regions before the extensive water-supply systems were installed. In 1973, a sand dam equipped with perforated metal pipes covered in gravel and sand provided water for livestock in underutilized South-Western rangelands (Sivils and Brock, 1981). Later, Bleich and Weaver (1983) proposed improvements in efficiency, dependability, simplicity as well as decreased installation and maintenance costs.

From the impressive number of dams in Kenya, not all these sand dams have resulted in increased access to water for local communities. Following the intensification of sand dams construction, several sand dams have been reported to have failed to supply water as expected (Olufisayo et al. 2010; De Trincheria et al. 2015; Nissen-Petersen 2006, 2011; Ertsen and Hut 2009). Nissen-Petersen (2006), a key stakeholder in sand dams’ development, admitted that 9 out of 10 recently constructed dams were not functional in South Eastern Kenya although this study did not confirm it. Each sand dam costs an estimated United States Dollar (USD) 5000 (Ertsen and Hut 2009) to 10,000 (Excellent Development 2016), resulting in wastage of valuable resources when dams fail. Dam failures have been blamed on poor site identification, environmental degradation and poor construction procedures.

When defining those factors that must play a role towards increasing access to water from sand dams, developers and communities used logical reasoning. Their river networks contained only coarse sand, beneath which they found water whose quantity they sought to increase. They reasoned the catchment soil textural characteristic should be able to yield coarse sand and recommended those high in sand content. Gijsbertsen (2007) supported this reasoning, and studied suitable catchments as being those whose riverbeds were full of coarse sand. Nissen-Petersen (2013) recommended catchments with stones and rocky hills that were a source of coarse sand into the riverbed during rainfall storms. The storms also flushed away clay and silt particles further downstream leaving the coarse sand behind. Presence of rock outcrops implied there was more of it underneath which acted as an impermeable layer that held water in place. Developers needed the land to be sloping to aid the rainfall delivering the coarse sand and, therefore, realized flat and very gently sloping land should be avoided. Gijsbertsen (2007) showed indeed, very gentle slopes resulted in river beds with fine-textured sediments.

Developers also needed river banks that would not collapse and fill the sand dam with the soil of undesirable textural characteristics and avoid re-routing. Besides, it was on these riverbanks that the masonry walls would be anchored in a trench. It was found better to erect sand dams on sections of the riverbed that contained more subsurface water, as those sections were last to dry up. Riverbed sections that had water indicating vegetation were also favoured (Nissen-Petersen 2000). Thus, rainfall, soil, topography, lithology, riverbanks, water indicating vegetation, sections of the riverbed where sub-surface water lasted at least one month after it got depleted in other parts were all included in the site selection criteria for sand dams. Finally, developers recommended building sand dams across riverbeds that were less than 25 m wide (Nissen-Petersen 2006) as more was found to be too expensive relative to annual household incomes estimated at USD 600 (MOA 2014). To make sure silt and clay particles were not retained in the sand dam, it was recommended to raise the dam height in stages equivalent to the expected quantity of coarse sand.

Despite these clear indicators, finding suitable sites for constructing sand dams is not straightforward. Catchments may change. A catchment with adequate amounts of sand to fill a sand dam within 2 years could have lower yields after soil and water conservation measures like terracing land (Tiffen et al. 1994; Wang et al. 2013) or produce sediments of different texture (Ouyang and Bartholic 1997a, 1997b). Gully erosion – common in Kenya’s ASALs – may result in sediments with undesirable textures (Nissen-Petersen 2011). A gentle slope may still yield more coarse sand than a steep slope in different catchments with similar soil and rainfall characteristics due to interactions between geology or human occupation. In general, as land use and land cover changes may affect textural characteristics of sediments getting into the sand dams, the current existence of coarse sand in a riverbed to determine whether a catchment yields adequate sand may be misleading. Even catchments with low quantities of sand will have managed to fill the riverbed with sand in longer periods.

The slope of the land may not be uniform and can vary greatly within the same catchment, which results in both erosion and deposition occurring concurrently (Alatorre et al. 2012). The slope of a catchment can be artificially modified through human activities (Tiffen et al. 1994). An impermeable rock or clay layer providing an impermeable base beneath the sand in a riverbed is not easy to identify, as its extent cannot be established without thorough and detailed geological assessment, which was the reason for Gijsbertsen (2007) to ignore this property of a site. The rock may be porous as well. Furthermore, as has been reported by Woodhouse (1991), tectonic movements can create faulting in geology causing impermeable rocks to leak, possibly resulting in sand dams losing their water altogether. Water indicating vegetation may no longer be at sites, as trees may have been cut down for timber, firewood and charcoal.

The rainfall amount adequate for water supply from sand dams is poorly defined as well. Adham et al. (2016) recommended regions with more than 200 mm. Mati et al. (2007) recommended rainfall/Evapotranspiration (ET) ratio (moisture index) of < 60% and presence of ephemeral rivers without defining a lower limit. Beimers et al (2001) mentioned a (high) range of 700–900 mm, as most ASALs receive less than 500 mm. Rainfall amount, intensity and distribution are highly dynamic in ASALs, which make recharging sand dams unpredictable as well. Spatial rainfall variability within the same catchment implies a dry section of a catchment could receive runoff from upstream, complicating defining minimum rainfall amounts for successful water supply from a dam. Rainfall amounts in ASALs seem to be declining, implying a reduction in runoff and sediments (Khisa et al. 2014; Wamari et al. 2012; Shisanya et al. 2011; FEWSNET 2010; Patricia 2010; Njiru 2009). Getting sites that meet all the listed conditions (Table 1) is difficult. Furthermore, the factors themselves are complex and dynamic. It is expected that existing sand dams, both successful and unsuccessful ones show complex configurations of characteristics, with their ability to supply water varying as much as these characteristics.

To have a grip on dam performance in relation to the factors, this research has assessed some of the stated environmental factors individually to identify their contribution to water access from sand dams. It explores a set of environmental characteristics that are considered to contribute to sand dams’ ability to increase water supply from a riverbed. Jamali et al (2013, 2014) considered abiotic factors such as slope, geology and soil within GIS including Topographic Wetness Index (TWI) values and stratigraphic information to identify suitable sites for subsurface dams. Encouraging results were obtained in regions with humid climates and limited natural water storage capacities. Adham et al. (2016) confirmed the use of these factors in its review of 48 studies on siting of rainwater harvesting structures in ASAL regions. In addition, findings between studies were also based on existing regional biophysical features, which implies that results from one region may not be applicable elsewhere (Adham et al (2016).

Environmental factors do not function in isolation; their combined effects may not be explained by assessing their individual impacts. As a response to interactions between environmental factors, the concept of AEZ may provide a useful guide for siting sand dams. Agro-ecological zones are regions that have exhibited homogenous climate, precipitation, physiographic factors, soils, vegetation and animal species. Key elements in defining an AEZ are growing period, temperature regime and soil type. The Food and Agriculture Organization and the International Institute for Applied System Analysis have developed maps of AEZs that considered interactions of similar factors and their influence on rain-fed crop yields (Fischer et al. 2002; Gunther et al. 2012). AEZ have been found to be reasonably good indicators of natural resources endowment and agriculture potential and in the formulation of strategies in soil conservation and fertility management, water harvesting, forest management and livestock development (FAO 2003).

Each AEZ has a specific range of potentials and constraints for land use (Berkat and Tazi, 2006; Sombroek et al. 1982). Kenyan agro-ecological zones were first outlined in Sombroek et al. (1982) and later revised by Jaetzold et al. (2006). The AEZ groups are temperature belts defined according to the maximum temperature limits, within which the major crops (maize, sunflower and cotton for Kitui County) in Kenya can flourish. Kitui county mainly falls in the transitional class 4 in the maize – sunflower upper midland (UM4) zone and marginal cotton zone in the lower midlands (LM4) (Jaetzold et al. 2006). Each unit has a different moisture index, which is the ratio of average annual rainfall for an area to average annual potential evapotranspiration in percentage form (Sombroek et al. 1982). In addition, rivers tend to converge and enlarge downstream increasing discharge. Downstream also have more subsurface flow entering the streams. To assess if this could have any correlation to dams ability to supply water during the drought periods, stream orders as defined by Strahler (1957) were included in the assessment as well.

With changes in environmental, anthropogenic, technological changes, even AEZs are not stable. A shift from high potential to low potential AEZs is attributable to climate change. In Kenya’s lowland zones (including Kitui County), there has been an increase of 0.5 ^°C in temperature (Jaetzold et al. 2006), which negatively affects the moisture index and lowers the potential to meet growth requirements for crops. Tiva catchment has moisture indices ranging from 15–65% across four different types of AEZs (semi-humid, semi-humid to semi-arid, semi-arid and arid) (Table 2). Tiva’s AEZs were analyzed to ascertain whether the interaction of factors that define the AEZs have any applicability in siting sand dams and whether the identified shifts in AEZs have any relevance on the ability of sand dams to supply moisture during the drought periods.

The study site was Tiva Catchment found within Kitui County. The identification of this study site was preceded by a reconnaissance survey in the counties of Kajiado, Machakos and Kitui. Kitui county was preferred for this study because unlike Machakos and Kajiado, it was less affected by sand mining. Tiva river catchment in Kitui County lies between − 1.66 and − 2.67 S and 37.75 and 38.67 E degrees, respectively (Fig. 2a, b). The county ranges between 400 and 1800 m above sea level (Lasage 2015), with the western and central parts characterized by hilly ridges separated by wide, low lying areas. Most of the land area is within the Tana River drainage basin. Only a narrow strip along the south and southwest border drains to the Athi River. In 2009, Kitui County had a population of just over one million people, with Akamba people dominating. Its main economic activities are subsistence agriculture, beekeeping, small-scale businesses, dairy farming, and ecotourism. Agricultural products from the study area comprise of fruits (mangoes, pawpaws, watermelons), maize, cowpeas, beans, pigeon peas, lentils and livestock (cattle, sheep, goats and chicken). Donkeys and oxen are extensively used for draught power, fetching water, transport and cultivation. Other economic activities in the county include charcoal burning, brick making and basket weaving (Ngigi 2003).

This study assessed 116 sand dams in October 2016 with 89 being within Tiva catchment, while the rest mapped just outside of the catchment boundary. October is generally the driest month of the inter-seasonal drought between long and short rains. The 89 sand dams were about 30% of the estimated 300–400 sand dams within Tiva river catchment, a number considered representative according to Mugenda and Mugenda (2003) and Borg and Gall (2003). A questionnaire was filled through interviewing sand dams’ beneficiaries and sand dams’ management committees whose coverage also considered gender balance. On each dam site, three dam beneficiaries were identified and interviewed. Data collected using the questionnaire at each dam site included socio-economic data, community water demand, the extent to which the sand dams met community water requirements, the time community stopped obtaining water from the sand dams and presence of sub-surface water for at least one month after the rains at the site where the sand dam was constructed. The information gathered through the questionnaire was clarified through observations and measurements in the field. The next sand dam was identified through snowball method where those interviewed directed the survey team to the next site. The geographical locations of dams were taken using a global positioning system (GPS) device and mapped onto the Kitui county map using geographic information system (GIS) software ArcMap. Geo-referenced maps of Kitui for factors that determine sand dams’ performance were acquired in form of raster images from the Regional Centre for Mapping of Resources for Development, Nairobi (RCMRD) and the Survey of Kenya. They included soil type, rainfall, digital elevation model from which catchment topography and stream orders were generated, geology/lithology and AEZ of 1982 and 2006. Details for the acquisition of each of the factors are listed in Table 3.

Data on socio-economic factors within Tiva catchment and sand dams’ ability to supply water from the questionnaire were analyzed using Statistical Package for the Social Sciences (SPSS) 22. Descriptive statistics such as frequencies, mean and percentages were used to describe population characteristics of user communities of sand dams. The raster images for soil, rainfall, topography, geology/lithology and stream orders were analyzed through creating polygons matching the different ranges. This was followed by overlaying the sampled sand dams whose ability to bridge water access gap had already been determined during the field survey. Interaction of these factors was analyzed with AEZ developed by Food and Agriculture Organization (FAO) and International Institute for Applied Systems Analysis (IIASA) (Sombroek et al. 1982) and revised by Jaetzold et al. (2006).

Projects for improving access to water in the ASAL regions such as Tiva catchment are perceived and expected to be drought relief interventions that would effectively deal with water-related problems. However, quite some interventions have failed to supply water as expected in Tiva, which remains a water-scarce region. Many of the sand dams constructed in Tiva catchment did not increase access to water during drought periods. This could be attributed to poor site identification, as Non-Governmental Organizations (NGOs) and government agencies involved in sand dams’ development may lack adequate guidelines for suitable site identification. At the same time, this study has demonstrated that existing criteria for identifying suitable sites for sand dams are insufficient. This suggests that proper site identification is rather complicated, even when strictly adhering to existing criteria. A suitable site requires more than meeting current pre-defined environmental factors and site characteristics. Some of the soil characteristics, topography and geology may not be useful factors at all when identifying suitable sites for constructing sand dams – at least that is what our results from Tiva catchment suggest. Rainfall amount, percentage of clay in soils, presence of rock outcrops, water indicating vegetation, agro-ecological zones and stream orders suggested some degree of predictability for sand dams’ ability to supply water in Tiva catchment. However, predictability remained rather low for most of the environmental factors, as some dams did not contain water in regions considered to be most suited. Using AEZs as guidance and stream orders may not provide all the answers to explain sand dam success, but would provide a basis for a more concrete plan that sand dam stakeholders can build on. Factors for success and failure should still be studied and documented, whereas sites where success factors for utilization of sand dams as water resources do not exist in their entity should be avoided.

The assessment we present also provides suggestions for which areas to avoid. Construction of sand dams in regions dominated by vertisols and ferralsols, areas receiving less than 600 mm of rainfall, AEZ of potential below LM4, stream orders 1, 2 and 3, should be avoided, unless more definitive identification criteria are provided. Likewise, dams without water in regions receiving more than 600 mm rainfall or in AEZs over LM4 and stream orders 4 and 5, should be studied to identify which sites to avoid. With this knowledge, suitable sites for construction of sand dams in regions with less than 600 mm, AEZ potential below LM4 and stream orders 1–3 could be identified. These four categories hold the answer to sand dams’ inability to supply water during the drought period in Tiva catchment in regions of both high and low potential and develop updated site identification criteria.

Targeted recharge of aquifers should be continued, among others through sand dams or other approaches. It is likely that some aquifers along the riverbeds can benefit from recharging using sand dams, but cannot be identified anymore, for example because water indicating trees have been felled. Past satellite images could be analyzed for such sites. Presence of rock outcrops at a site is not enough to guide to determine whether it would hold water during a drought. A qualified geologist should be called to ascertain whether a site will be permeable. If found to be permeable, sites can either be treated or should be abandoned. In regions where sand dams are found to be inappropriate, other water provision technologies could be identified and implemented.

A map for appropriate implementation of sand dam technology could be prepared for areas like Tiva catchment, using existing data to minimize guess work that may lead to more technology failures and wastage of resources. After all, costs of putting up sand dams that failed are high. With one sand dam costing an estimated USD 5000, in Tiva catchment an estimated USD 132,250 (23/100 × 115 × 5000) was lost on those dams that supplied water for just one week which can be considered as a complete failure. Alternative technologies, such as plastic-lined ponds or tanks, could have resulted in more access to water for communities. Plastic-lined ponds are cheap, but can only target individual farmers while water tanks can target schools and larger households. The challenge with these other technologies is that they do not readily render themselves for communal water access. Sand dams have probably gained popularity because they do build on community efforts, which is why sand dams may continue to be one of the more useful technologies to provide rural communities with adequate water, now and in future.

Concerning this future, our study suggests that increases in temperature through climate change would not necessarily adversely increase evaporation from sand dams nor effect their success. In regions where sand dams successfully supply water during the drought period, they could be an important climate change mitigation technology. Constructing those dams, however, should not be based on simple criteria that are applied without studying the environmental conditions in the depth that is required. There is a need to go back to the drawing board for sand dams to become the universally successful climate-mitigation technology that so many think they already are.